TUSKEGEE AND THE HEALTH OF BLACK MEN

Abstract

JEL Codes: I14, O15 For forty years, the Tuskegee Study of Untreated Syphilis in the Negro Male passively monitored hundreds of adult black males with syphilis despite the availability of effective treatment. The study's methods have become synonymous with exploitation and mistreatment by the medical profession. To identify the study's effects on the behavior and health of older black men, we use an interacted difference-in-difference-in-differences model, comparing older black men to other demographic groups, before and after the Tuskegee revelation, in varying proximity to the study's victims. We find that the disclosure of the study in 1972 is correlated with increases in medical mistrust and mortality and decreases in both outpatient and inpatient physician interactions for older black men. Our estimates imply life expectancy at age 45 for black men fell by up to 1.5 years in response to the disclosure, accounting for approximately 35% of the 1980 life expectancy gap between black and white men and 25% of the gap between black men and women.

Figure I. Black-White Mortality Differences by Age and Sex

Notes: The data are from the CDC compressed mortality files and represent the black-white difference in age-specific mortality rates. Each mortality rate is calculated by dividing the number of deaths in the relevant population by the at-risk population (in thousands). The solid (blue) line represents the difference for males, and the dotted (red) line represents the difference for females. The vertical line represents the year “The Tuskegee Study of Untreated Syphilis in the Negro Male” was disclosed. For additional figures, including plots of all other age-specific mortality rates and South only, see the Appendix.

Figure II. Proximity Measures

Notes: Figure 2 plots proximity measures. Panel (A) represents distance from Macon County from the centroid of each SEA calculated using ArcGIS proximity tools. Panel (B) represents the fraction of black migrants from Alabama by SEA calculated using the 1940 census. Darker tones reflect closer proximity to Tuskegee or a higher fraction of Alabama migrants using 20 natural breaks. The white open circle in Panel (A) demarcates approximately 1000 kilometer radius from Macon County. The white closed circle in Panel (A) and black closed circle in Panel (B) represent Macon County.

Figure III. Difference-in-Differences Coefficients and Geographic Gradients for Chronic Cause Mortality

Notes: These figures plot β coefficients from equation (1) estimated on various demographic group samples (all-male, all-black, all-white and all-female) as indicated by the figure headings and further subdivided into distance bins from Tuskegee. The outcome variable for each regression is the log of age-adjusted chronic mortality. Panels (A) and (B) plot β coefficients on black*male*post in the all-male and all-black sample, respectively. Panels (C) and (D) plot β coefficients on white*male*post and black*female*post in the all-white and all-female sample, respectively. Panels (E) and (F) plot the same coefficients as in Panels (A) and (B), however, the time period of the analysis is restricted to the pre-1974 period with all observations after 1969 constituting the “post” period. SEAs were grouped into K bins of 150 kilometers of distance from Macon County, Alabama each bin yielding a β^k. Bins in the middle of the country with less than 2,500 black men in 1970 were combined, and confidence intervals that were greater than 0.6 in absolute value were omitted so that the scale could be the same for all figures. Regressions for a given bin were weighted by the relevant SEA population (e.g. 45–74 year old males for the all-male sample), and the size of the bubble reflects the total weight of the respective treated population within each distance bin. Each β^k coefficient is indicated by a bubble and the dashed vertical lines connected to the bubble denote its 95% confidence interval. The thick (blue) line depicts the average β coefficient across the entire subsample. This average estimate and its 95% confidence interval are provided in the figure subheading. The line of best fit between the β coefficients and distance from Tuskegee is drawn as a dashed (red) line. Data are from the CDC compressed mortality files and cover the period 1968 to 1987 unless otherwise indicated. Rates are constructed biennially, and post is an indicator equal to 1 in the years following 1972/1973 for Panels (A) through (D). Placebo post is an indicator equal to 1 in the years 1970/1971 and covers the period 1968 to 1973. See main text and Appendix for further details on the data source and variable construction.

Figure IV. Event studies

Notes: Event study coefficients from versions of equation (4). Panel (A) plots the β₁ coefficients on the interaction of proximity, an indicator for black, an indicator for male, and a year indicator. Panel (B) plots β₂ coefficients, the interaction of proximity, an indicator for male, and a year indicator. Panel (C) plots β₃ coefficients, the interaction of proximity, an indicator for black, and a year indicator. The dependent variable for Panels (A) through (C) is the number of outpatient visits within the last 12 months. The sample includes non-veteran black and white men and women ages 45–74 and covers the period 1969–1977. Data are from the harmonized version of the National Health Interview Survey (NHIS) available from IPUMS and merged with restricted identifiers for use in the RDC. Panels (D) – (F) plot the β₁, β₂, and β₃ coefficients for the outcome of (log) age-adjusted chronic mortality constructed biennially. Mortality data are from the compressed mortality files from the CDC and cover the period 1968 to 1987. The unit of observation is the state economic area (SEA), and the sample includes black and white men and women ages 45–74 who died in the United States. Each estimate is relative to either 1972 (vertical dashed line in Panels (A) – (C)) or 1972/1973 (vertical dashed line in Panels (D) – (F)). See Table I notes, text and Appendix for further details. ** p<0.05.

Figure V. Permutation Test

Notes: Frequency of the “false” β₁ coefficient estimated using distance from every other state (exclusive of Alabama) or SEA (exclusive of the one containing Macon County) in the sample and estimating equation (2) in Panel (A) or equation (3) in Panel (B). The vertical line denotes β₁ from baseline estimates using the true treatment distance (to Macon County, Alabama) as reported in Table I.